Prediction of electrical resistivity of steel using artificial neural network

Abstract

Electrical resistivity of commercially produced plain carbon manganese steel has been experimentally measured at room temperature (28–30°C) using four-probe method. Resulting data were used to generate both regression based and artificial neural network-based models for prediction of electrical resistivity from the chemical composition of steel. It was found that both models were capable of predicting the resistivity within ±5% error band. Analysis of data also indicated carbon to be the most influential element to increase resistivity followed by manganese and silicon. A comprehensive literature review indicates no such advanced resistivity prediction model is available in the public literature for commercially produced steel with wide variation in carbon content (0.03 0.85 wt-%), manganese content (0.35–1.50 wt-%) and silicon content (0.015–0.90 wt-%).